Paul Novosad
@paulnovosad
Followers
15,028
Following
588
Media
569
Statuses
6,970
econ prof @dartmouth , founder . r2: "a morass of disjointed streams of consciousness" 🤷
NH
Joined January 2012
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Olympics
• 3863746 Tweets
Venezuela
• 1138221 Tweets
Bitcoin
• 473774 Tweets
Christianity
• 465499 Tweets
Islam
• 428271 Tweets
Last Supper
• 377452 Tweets
Hezbollah
• 287496 Tweets
Majdal Shams
• 135008 Tweets
Sanchez
• 113000 Tweets
The Villages
• 106110 Tweets
Greece
• 89520 Tweets
لبنان
• 84467 Tweets
Nadal
• 83689 Tweets
Druze
• 82261 Tweets
Lebanon
• 75814 Tweets
مجدل شمس
• 69626 Tweets
金メダル
• 61070 Tweets
Jazz
• 48833 Tweets
C Spire
• 44289 Tweets
Alcaraz
• 40528 Tweets
Medina
• 33404 Tweets
YO SIGO A DENISSE PORQUE
• 31172 Tweets
Antoine Dupont
• 22802 Tweets
Gary Gensler
• 19526 Tweets
Maresca
• 16920 Tweets
Giannis
• 16628 Tweets
#アイドルへモーニングコール
• 16035 Tweets
Suns
• 14046 Tweets
Shai
• 12447 Tweets
Pinned Tweet
📣 New working paper on residential segregation in India. We’ve been working for 5 years on this.
8 facts about residential segregation in India, from new administrative data. The situation is not great 🧵 1/N
196
2K
6K
I knew scientists were territorial but this is next level.
Remember Katalin Karikó, the unsung hero of mRNA vaccines? Her advisor Robert Suhadolnik tried to have her deported when she accepted a competing offer in another lab.
112
3K
10K
Dartmouth's student government held an open "no confidence" vote on the president's handling of campus protests. It passed (no confidence) 13-2 (with 3 abstensions).
Then they repeated with a secret ballot. It failed 8-9 (2 abstensions).
1/2
78
483
3K
What gets you into a top college is being above some moderate competence bar, and then, from like 8th grade or earlier, organizing your life around getting into a top college.
Very few 8th graders have this level of motivation, so we mainly select on parent characteristics.
43
106
3K
Wow, quite the story. 👇🏻👇🏻
Founding Arxiv is apparently not good for your academic career.
Good evidence that improving how science works is insufficiently incentivized.
14
417
2K
🤷🤷♀️New data: SHRUG 2 is out!!
@devdatalab
has been working on this for two years, a HUGE update to India’s coolest data platform:
1. Maps of *every* 2011 town and village, with ids
2. All data at every geography (villages, districts, ACs, etc)...
🧵1/N
35
607
2K
Re-upping this story about how Katalin Karikó's advisor threatened to have her deported if she moved labs.
I knew scientists were territorial but this is next level.
Remember Katalin Karikó, the unsung hero of mRNA vaccines? Her advisor Robert Suhadolnik tried to have her deported when she accepted a competing offer in another lab.
112
3K
10K
24
513
2K
Behavioral econ is cool, but once we admitted people are irrational, it seems wrong to end up with 50x more work on "we must help people with tiny nudges," vs. "Oh shit, firms must be taking advantage of people like crazy with this stuff."
40
160
2K
I've lately reviewed many papers that *feel* p-hacked.
Some clues: results implausibly large, operate at the wrong unit of observation, too perfect, etc
A couple of specific author teams are doing this very systematically.
I think this is a canary in the coal mine. 🧵 1/N
41
161
2K
Is there a theoretical basis for being comfortable with charging men more for car insurance? It's statistical discrimination, it's clear why firms do this, and it seems fine with me.
But do we have a theory for deciding which forms of statistical discrimination we are fine with?
218
196
2K
I've been telling people about this paper for a while, it's sooooo neat.
Prayers for rain only "work" in places where the probability of rain is increasing in the length of the dry spell ---> so these places have more religious belief.
24
321
2K
When you hear how many people say or support something, you always have to understand the social dynamics in which these "votes" take place.
Many institutional and social structures are suppressing people's true opinions. 2/2
13
83
1K
RCT reducing leisure screen time among adults and adolescents to *<3 hours per week*. Compliance was high — they exchanged smartphones for flip phones.
Results:
- 45 mins more physical activity per day
- Better mood and well-being (adults)
- Fewer behavioral issues (kids)
13
249
1K
🚨Please stop calling this brain drain.
When there are high overseas returns to education, *more people get educated*.
Think about all the folk working their asses off for this exam, and then *staying in India*.
High returns for migrants are good, they build human capital. 1/N
Out of the 1,000 top scorers on the entrance exam for the Indian Institutes of Technology, 36% have migrated eight years later (primarily to the US).
Out of the top 100, 62%.
Out of the top 10, 90%.
136
951
4K
38
170
1K
📣📣 New data alert: we
@devdatalab
are releasing open village and town maps for all of India. Find them at the top link here:
We wrote a post with some more details about the maps:
@thesamasher
@tobylunt
1/5
12
284
1K
I'm just so sad that so many progressives have abandoned the ideal that we shouldn't treat people differently based on their skin color.
I'm baffled people are willing to defend this. I don't think this is getting us to a better place.
53
62
1K
We are letting our allies down. Exporting 100% of our AstraZeneca is a no-brainer at this point.
11
159
975
A 🧵 on our work on US mortality change, just out in AEJ:App, with
@thesamasher
and
@charlierafkin
.
We ask: how concentrated is the U.S. pre-Covid mortality crisis? Is everyone doing a little worse, or is a small subset doing catastrophically worse?
The graph is a spoiler 1/N
19
279
968
Good wikipedia nugget about today's nobel winner Carolyn Bertozzi.
18
139
975
Reasons you are probably overestimating the probability that Twitter dies:
1. Platforms die when their audience leaves. Twitter audience is bigger than ever.
2. Platforms survive longer than you think. MySpace, Tumblr, Flickr, even AOL are still alive.
1/N
82
157
938
Our study on excess mortality in India is out:
Excess deaths close to 3 million.
1. Far higher than official totals.
2. Implies global COVID death count off by >2 million (and more given undercounting elsewhere)
Short 🧵
65
537
936
Pfizer is creating hundreds of billions of dollars of social value, and stockholders have earned about a 5% return since Jan 2020.
29
76
927
I like these article highlights — Elsevier showing they can still add value to the publication process.
11
92
793
What a graph! High-income parents teach their children independence, low-income parents teach obedience.
Source: WVS via Acemoglu
Theory: parents teach traits that are optimal for the expected work evironment of their kids— limiting upward mobility.
6
205
763
U.S. working hours per year are lower than they have ever been.
The thing that is demanding your constant, full attention and dysregulating your nervous system is your iPhone.
Capitalism demands our full attention, working constantly. It doesn’t give us room to actually be human, feel feelings, or take time to breathe. It makes it so our nervous system is always dysregulated. This isn’t sustainable, dude. We can’t keep living this way.
347
1K
4K
16
95
722
Fact 1: India is very segregated.
Urban places are about as segregated as rural places, for Scheduled Castes. For Muslims, segregation is worse in cities. The graph shows the Dissimilarity Index for cities and subdistricts.
11
210
639
I find it wild that people out there think that essays are more egalitarian than exam scores
Wow: “He started by editing college essays from his Yale dorm room for $50 an hour but now charges the parents of his company’s 190 clients $120,000 a year to help them create a narrative he believes will appeal to college-admissions officers.”
21
13
91
16
63
590
Fact 2: Scheduled Castes and Muslims are about as segregated as Black people in U.S. cities. Note this dissimilarity graph is slightly different from the prior, b/c we limit to cities >100k to match U.S. Census definitions. 5/N
6
180
577
I'm surprised I haven't seen this paper until now (by
@mikekofoed
).
Clean RCT of zoom- vs. in-person-school: 38 sections, same class (intro econ), same time slot, same instructor. Zoom students did 0.2 SD worse.
17
169
549
TurboTax is an influential lobby in favor of keeping taxes difficult, to drive demand for their product.
How much would it cost a philanthropist to fund an open-source alternative to drive TurboTax out of business and open to door to tax simplification?
23
39
547
Making your paper general interest by taking the country name out of the abstract
7
28
522
Michael Kremer could be the first economist to win two nobel prizes, the first for RCTs in development, the second for advance market commitments for vaccines.
11
76
506
In the long run, I think we fix this by requiring transparent data and code replication sites before submission. But this feels a long way distant, in between I think a lot of fraud is going uncaught... N/N
16
24
504
Nice to see our paper on 30-year U.S. mortality change covered in Matt Yglesias' newsletter today with a very clear summary.
The U.S. is not experiencing a widespread malaise, it's a concentrated catastrophe among the least educated.
10
128
506
This is really neat: Community Notes work because they estimate users' political preferences and upweight notes that are liked by users of all stripes.
I wonder if other parts of our democracy could be improved with mechanisms like this.
8
72
497
Fact 3: Muslims are more likely to live in highly segregated neighborhoods.
26% of urban Muslims live in neighborhoods that are >80% Muslim.
17% of urban SCs live in neighborhoods that are >80% SC.
Numbers in rural areas are similar. 6/N
14
143
479
Fact 5: Public services in cities are less likely to be found in neighborhoods with many SCs and Muslims.
A 100% Muslim neighborhood is only half as likely to have a secondary school as a neighborhood with no Muslims. 10/N
12
163
480
Our current peer review system isn't robust to blatant cheating.
The Ariely types uncovered based on different fonts in Excel sheets are probably outliers — the successful frauds are usually better at covering their tracks. 3/N
4
15
468
People dunking on this too quickly, "correlation is not causation", "we don't vote count studies", etc.
*Not every important question can be answered with a perfect RCT*
For decades, the cigarette-cancer link was similarly correlational, based on case studies, etc. 🧵 1/N
A review in Nature, by
@candice_odgers
, asserts that I have mistaken correlation for causation and that “there is no evidence that using these platforms is rewiring children’s brains or driving an epidemic of mental illness.” Both of these assertions are untrue.
298
2K
8K
14
65
472
A+ Intro Econ content right here.
7
40
453
Proposal for a new residential building in New York spanning the East River
48
28
431
A key discovery of the RCT revolution in development was that most programs did not work as they were supposed to.
But when I get asked to referee papers these days, it's always about a low-cost intervention with transformational effects.
I don't know what to make of it. 1/n
17
58
432
It's mainly: did parents steer the kid's energy into activities that build a college CV.
Some of these activities involve real human capital accumulation, like Russian math.
Too many of them, like starting NGOs, writing research papers, stints overseas, are mainly signaling.
3
7
425
On referee advice for our null bias paper, we formally tested for publication bias.
Good news: We were right about publication bias: Random judge assignment studies that don't find bias are *30x less likely* to be published.
Bad news: Our paper is never getting published.
2
67
408
The elite college "meritocracy" in the U.S. rests on the idea that you can tell who the "meritorious" people are by age 17.
It's ludicrous — the only thing you are only learning about is their parents.
16
40
405
Is the reason "statistical discrimination is only bad if it's against groups that were discriminated against in the past"?
This is the best I can come up with, but it feels circular and dissatisfyingly vague — like how much past discrimination for how much present?
16
4
392
Fact 4: Cities replicate the social environments of their hinterlands. Districts with segregated villages have segregated cities. 7/N
4
68
375
Uh-oh — virtually all major RCTs funded by an NIH institute (NHLBI) before 2000 were false positives. Once hypothesis preregistration is required in 2000, everything becomes a null.
Via
@pmarca
/
@RichardHanania
.
6
84
364
It's disturbing, because these authors are not far off from getting away with it.
A little more care for the details and these papers are going to publish in top journals and be indistinguishable from papers that are learning true things. 2/N
3
6
353
We built a national neighborhood-level dataset covering all India, 2011–13.
It’s super local. A neighborhood = ~700 people, 1.5m in the country.
Data are from ~2012, this is about historical patterns, not the current govt.
3/N
3
40
342
The best thing about writing in economics is that you can get the entire argument of the paper by reading the introduction.
Abstract: TL;DR
Intro: short version
Rest of paper: long version
Other social scientists: please adopt this approach.
12
29
349
I'm so glad somebody finally wrote this paper!
My referee reports on close election politician characteristics papers will be much shorter.
5
45
354
Imagine if the US government froze bank accounts of people who made donations in support of the BLM protests last year.
This shouldn’t happen in a free society.
Deputy PM Chrystia Freeland: "The names of both individuals and entities as well as crypto wallets have been shared by the RCMP with financial institutions and accounts have been frozen and more accounts will be frozen."
7K
4K
8K
18
31
332
"Men who lose access to Medicaid eligibility are 14% more likely to be incarcerated in the subsequent two years relative to a matched comparison group. The effects are entirely driven by men with mental health histories."
by
@elisajacome
7
93
331
The Economist's daily chart is from our paper on mining and criminal politicians in India.
More mining rents — worse politicians, worse behavior in office.
4
94
337
If you work for a mission-driven org, one of the best things you can do is hire people right out of undergrad.
Students are setting down life paths, Wall Street and Facebook are knocking on their doors, and great NGOs are like "contact us when you have 3–5 years of experience"
5
33
329
The best job market candidates signal both interest and competence by including a structural model in the cover letter.
6
7
327
This is joint work with
@thesamasher
@kritarthjha
@aadukia
@brandonjoeltan
Summary:
Media fact sheet:
Paper:
Let’s begin 2/N 👇
3
40
310
I had not read the emails before. This is fucking insane. What a blatant violation of academic integrity.
How are these people in positions of power?
(Rhetorical question — our institutions reward this shit, that's why these people are in power)
6
37
319
The existence of segregation is not surprising to people who study and spend time in Indian cities.
Some of these descriptive facts have been noted by
@RaphaelSusewind
,
@nav_bharathi
,
@deepak_malghan
, among others. 8/N
1
39
308
For most of history, and in most places still, people openly discriminate on race/ethnicity, think different ethnic groups have different essential characteristics.
Abandoning these ideas in the west was one of the triumphs of the 20th century.
What a waste to give up on this.
4
14
312
My heart sinks when I get asked to review a paper by one of these teams. (Ok fine, this happens when I get asked to review any paper.)
Decline and make it someone else's problem? Whisper to the editor — implicitly evaluating based on author identity rather than content? 4/N
7
4
306
When we look at SCs, moderate SC neighborhoods are doing ok, but the most segregated neighborhoods again are less likely to have secondary schools. 11/N
1
58
302
The result is systematic: within cities, neighborhoods with high SC and Muslim shares have much worse public services—look at the rightmost red bar. The other services are in Figs 5–7 of the paper.
6
118
298
We are comparing SC, Muslim, and integrated neighborhoods, *within the same city*.
This kind of granular data has not been available before. If you look at geographic aggregates, like districts, you will find a different (misleading) story. 12/N
1
39
283
I’m not an expert on oil dependence, however I do know this — nearly everyone always underestimates supply elasticities.
If you raise the price of oil and restrict Russian natural gas, firms and the market will find a way to adjust, and will do so better than you expect them to.
11
38
291
Top5 econ journals have too much power. They use it to extort authors in unproductive ways.
It is easy to reduce their status.
Just stop using the phrase "top 5" in reference letters. That's it. If you want a rank then say "top 10." Or talk about the importance of the work.
14
28
290
"Brain drain" is mostly a myth. It assumes the stock of educated people is fixed.
It isn't— when engineers are getting amazing international opportunities, *more people train to be engineers.*
2/N
8
26
287
@izakaminska
The graphs show that Uber's profit is small relative to revenue.
This has nothing to do with whether their business is or isn't zero sum.
Next, do grocery stores.
5
1
288
But across neighborhoods, the results are stark. Muslim neighborhoods are *much less likely* to have public secondary schools. 16/N
9
83
275
This is how the trends look once you hold ranks constant. From 1992–2018, most White Americans have been doing fine mortality-wise — but the least educated 10% have faced catastrophic mortality increases. 9/N
5
87
276
Once you look at school allocation *across* blocks/neighborhoods *within* towns, most of that advantage disappears.
For whatever reason, within cities and towns, the most segregated SC neighborhoods have the fewest secondary schools. 14/N
3
52
261
Fact 5: Kids are worse off in segregated neighborhoods.
Young people have over a full year less education in fully segregated SC and Muslim neighborhoods. The graph shows outcomes for 17–18 year olds. 20/N
1
65
253
But maybe groups choose to live together — does it matter?
Let’s look at service delivery in these neighborhoods, starting with secondary schools 👇 9/N
4
31
247
Interesting paper: "The experience of effortful thinking itself increases the ability to accumulate traditional human capital."
If you're on Twitter, you're getting the opposite treatment of this experiment.
6
59
246
A key takeaway from the Chetty/Deming/Friedman paper is that recruiters should obsess less with candidates from Ivy+ schools.
There are exceptional candidates at other places that are being overlooked, they have everything except the rich parents.
7
27
248
3. Twitter's main asset is the network — you. You are all still here, even the ones who left are still here.
4. The media class despises Elon Musk and is cheering for his demise. They are extremely biased observers.
2/N
7
17
238
100% we never lived up to this ideal and there is a lot of work left to do.
But these progressive neo-segregationist ideas are so short-sighted and divisive.
As if doubling down on tribalism could ever help us get past a history of tribalism.
7
8
243
Fact 6: Kids from *all social groups* are worse off in segregated neighborhoods.
The neighborhood effect explains about half of the group disadvantage. In predicting your education, where you live is just as important as your social group.
1
57
235
An R&R resubmission where there are 100 pages of responses to the referees, but nothing substantive in the paper is changed.
What a waste of time for everyone involved.
Sadly, this is my modal experience when refereeing R&Rs.
4
12
242
This graph shows school access vs. SC share at aggregate levels.
The story is positive for SCs: states, districts, and towns with more SCs all have more secondary schools, maybe because policies have targeted schools to high-SC regions.
But …
1
37
228
SHRUG has new data on banking in India! Opening date, village/town and branch name for 150,000 banks in India. Cool!!
Huge thanks to Sandhya Garg and Samarth Gupta for contributing this. (Cite them!)
Check out their paper on village banking expansion:
1
47
237
The takeaway is not: "more robustness checks."
The takeaway is: "authors cannot be trusted to run their own robustness checks."
We need fewer referee demands for robustness, and more replication by scholars who have no incentives to cheat.
How credible was the "credibility revolution"? How robust is empirical research in economics? We just replicated a year's worth of the American Economic Review & had economists predict robustness. Here's what we learned.
22
254
673
6
23
236
Few economists bite the bullet that if exporting is good for the countries sending out the goods, then importing must be bad for the countries bringing in the goods.
20
9
231
Very bad and fascinating. Paper link and abstract for those who can't tolerate cute video explainers.
@Eyal_Frank
and I show the sudden extinction of vultures in India due to the painkiller diclofenac led to a ~4% increase in all-cause mortality. Why? Econimate explains in video. PS: Working paper on SSRN. Being *revised* so comments welcome but code online only once accepted.
7
28
115
5
65
234
8. Mastodon isn't ready, it's badly designed, it's unfun.
It's like the table of contents email from the Review of Economic Studies. It's probably interesting and you would learn something if you read it, but it's not *that* interesting and you're not going to.
8
24
228
Broke: eliminate hard math classes for all, to help low-income students.
Woke: create more hard classes, and help low-income students get into them.
1
33
226
Does temperature affect judge decisions? These two articles are using literally the same strategy and the same dataset, but finding 3x difference in magnitudes.
All the data and code are public, it would be a great class/twitter project to dig into this.
Links in thread.
5
25
225
The difficulty of canceling a
@nytimes
subscription is so scummy and dishonest.
It can't be worth the reputational cost to the company for the few extra people who don't have the patience to unsubscribe.
17
6
224
We included Facebook's Relative Wealth index in the SHRUG, as it is one of the only granular measures of living standards in India since 2012.
It's under-used, read the thread and check it out -->
Meta is using Facebook and mobile phone data to produce super-granular household wealth estimates.
Here’s what you need to know about the Relative Wealth Index:
17
305
1K
3
36
224
I don't know, I feel like the NYT should use a higher standard for reporting on Covid studies than for reporting on nutrition studies.
This study might be correct, but there are also a lot of reasons to think it might not be the final word. 1/N
5
42
217